There is now substantial evidence to support earlier hopes that all night electroencephalographic sleep studies supplemented by psychological measurements, motor activity monitoring and clinical psychopharmacologic techniques will add an important dimension to the differential diagnosis and treatment of depression. In the context of the general hypothesis that the severity and type of affective disorder can be identified and the efficacy of classes of psychotropic drugs used in the treatment of depression can be tested with a classification derived from a constellation of sleep parameters, we plan to test the following in 200 consecutively admitted patients to the clinical research unit with a diagnosis of major depressive disorder (both primary and secondary depression) requiring hospitalization: 1. that the indices derived from sleep parameters together with other psychobiologic measurements can substantially improve treatment response prediction by delineating the psychobiologic profile of various depressive subtypes: 2. more specifically, that depressed patients with shortened REM latency are antidepressant drug responders and that a constellation of REM measures predicts response to tricyclic antidepressants with a high degree of probability; 3. that the sleep characteristics of psychotic depressives are significantly different from the sleep profile in nonpsychotic depressives and that differences can be used to improve the chances of treatment success; 4. that the clinical characteristics and treatment response are different in depressed patients with a normal REM latency from those with a short REM latency. In addition, we are interested in defining what are the sleep changes that might precede, coincide with, or even lag behind clinical relapse and remission in depressive states and whether certain sleep abnormalities (e.g., short REM latency) are present during periods of remission as psychobiologic "markers".